OBJECTIVES The aim of this study is to assess for 2012 the cost of diabetes from a payer perspective, based on the French health insurance database and using two different approaches (top down versus bottom-up). METHODS Using information about 60 millions of individuals from the general scheme insurance database (85% of the French population), we developed algorithms to identify all people who received care for each of 56 groups of diseases or medical events or treatments, which are frequent, severe and/or costly. Algorithms have been applied to each patient. For diabetes, we used ICD-10 diagnoses for long-term chronic diseases, reimbursement for anti-diabetic drugs. Costs of all reimbursed expenditures (outpatient/inpatient care, disability/sickness benefits) were extracted per individual. The top-down method allocated expenditure to each of the 56 diseases based on the average expenditure by disease calculated for individuals with only one disease. All expenditures were thereafter extrapolated to the whole population to fit national health account aggregates. For the bottom-up approach, diabetes expenditures were estimated by identifying finely in our database expenditure items which are partly or wholly directly related to diabetes according to expert judgment. RESULTS Based on the top-down approach, among the 146 billion euros of expenditures reimbursed by national health insurance (all insurance schemes) in 2012, 7.5 billion (5%) are attributable directly to diabetes. Expenditures for chronic renal insufficiency and cardio-vascular disease, frequent diabetes complications, have been assessed separately. Drugs and medical devices (3.8 billion €) represents more than 50% of the diabetes expenditures, other outpatient care 34% (2.5 billion), inpatient care 9% (700 millions) and disability/sickness benefits 7% (500 millions). Based on our bottom-up, around 8 billion would be directly attributable to diabetes. CONCLUSIONS Our study provides estimation of the cost of diabetes from a payer perspective, according to two different approaches but with concordant results.